class torch.nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True)
Conv2d
原文：https://www.cnblogs.com/liujianing/p/12262570.html
